Racial profiling: step aside. It seems facial profiling is the new hot topic. Late last week, the Wall Street Journal reported that law enforcement officers across the country would begin carrying hand-held facial-recognition devices in September, which brought issues of privacy to the forefront of the minds of many.
But are people rightfully worried?
The biometric device in question, known as MORIS, can identify people based on their faces, irises and fingerprints. But news about this product makes facial recognition sound easier than it actually is.
For example, MORIS cannot identify people blindly; it must have a database to draw from. The smaller a database is, the easier it is for the device to work, because the margin of error is less. And the technology isn’t advanced enough to be able to easily pick out a face in a crowd; something like a security camera at an airport could capture live footage, but would likely be unable to place terrorists in a group of hundreds or thousands.
The actual database in question is criminal in nature; the following graphic is the Wall Street Journal’s rendition of how the tool and the database work together:
This kind of technology could, however, have practical uses for a company with high security. Employees could be issued keycards which, when scanned at an entrance, then activate computer-vision software. This software would scan the face of the person at the door, validating identity and linking each key with an individual face. The proper combination of a person’s key and the corresponding face in the database would then unlock a door.
Interestingly enough, the term facial recognition is familiar to many, but the definition most people are familiar with doesn’t completely encapsulate exactly what kind of technology this is, and what it does.
In short: facial recognition requires a database to draw from in order to complete a task. But there does exist another technology, known as object recognition, which can find and identify an object within a series of images or videos, even if the objects are distorted or partially hidden.